An Emerging Field of Psychological Enquiry
People are increasingly using conversational AI for emotional support, mental-health information and as an adjunct to existing psychological care (American Psychological Association [APA], 2026).
Emerging evidence suggests potential therapeutic applications of generative-AI mental-health chatbots, although the evidence base remains developing and questions concerning effectiveness, safety and appropriate clinical integration require further investigation (Zhang et al., 2025).
The concept of a digital therapeutic alliance is also beginning to receive empirical and theoretical attention, particularly in relation to engagement, perceived empathy, therapeutic goals and the nature of the relational bond formed through digital interventions (Malouin-Lachance et al., 2025).
The integration of AI into psychological and healthcare contexts also requires attention to autonomy, transparency, safety, accountability and human oversight (World Health Organization [WHO], 2021).
Artificial intelligence is developing more rapidly than our understanding of its psychological consequences.
People are already using conversational AI to discuss grief, relationships, anxiety, loneliness, identity and other emotionally significant experiences. In the American Psychological Association’s 2026 survey of more than 1,200 practising psychologists, 77% reported having spoken with patients who had used AI for support, engagement or other purposes, while 35% reported patients using AI as an additional mental-health professional.
The question is therefore no longer simply whether people will bring AI into their psychological lives.
They already are.
The more important questions concern what happens when they do.
What We Know So Far
Research into digital mental-health interventions and AI-assisted psychological support is expanding.
Recent systematic reviews have examined generative-AI mental-health chatbots and reported promising findings alongside substantial variation in technologies, populations, research designs and levels of clinical validation. The evidence base remains considerably less mature than that supporting established psychotherapies.
Research is also beginning to examine whether interactions with digital systems can produce some of the experiences associated with a therapeutic alliance—including perceived empathy, engagement and collaboration—while raising important questions about what such a “digital therapeutic alliance” actually represents.
These developments suggest possibilities worth investigating.
They do not establish that generative AI can replace psychotherapy.
What We Still Need to Understand
Many of the most psychologically interesting questions remain open.
What happens when AI is used between psychotherapy sessions rather than instead of psychotherapy?
Can structured AI-assisted reflection help clients articulate experiences that are difficult to formulate alone?
Can it support reflection, emotional awareness and recognition of recurring relational patterns?
What material generated through AI interaction subsequently enters the therapeutic relationship?
Could discussing AI conversations with a therapist create useful opportunities for therapeutic exploration?
And equally importantly:
When might AI reinforce an inaccurate interpretation?
When does validation become over-validation?
Could repeated interaction encourage reassurance-seeking or dependency?
How do attachment patterns influence the way different people relate to conversational AI?
What happens when someone experiences an artificial system as understanding, caring or emotionally significant?
And how should therapists respond when these experiences appear within psychotherapy?
These questions require empirical investigation rather than assumption.
The Importance of Human Oversight
Safety must develop alongside innovation.
The World Health Organization has consistently emphasised ethics, human rights, accountability, transparency and appropriate governance in the development and implementation of AI for health. Its more recent work specifically highlights concerns surrounding generative AI and mental health, including emotional dependency, crisis response and the need for independent research into both short- and long-term outcomes.
AART-345 therefore approaches therapist involvement not simply as a feature of the model, but as an important area for investigation.
Does therapist-guided AI use differ psychologically from independent AI use?
And can supervision provide an additional layer of clinical scrutiny by helping professionals examine what occurs between the client, therapist and AI system?
These are potentially important research questions in their own right.
AART-345 as a Developing Research Framework
AART-345 is currently a developing clinical and research concept.
Its proposed 12-week structure provides an opportunity to investigate AI-assisted reflection within a clearly defined human therapeutic framework rather than allowing AI to function as an autonomous substitute for psychotherapy.
Potential research could examine changes across the three stages of the model—Reflect, Understand and Transform—and explore both therapeutic outcomes and the client’s evolving relationship with AI.
Research questions might include perceived usefulness of AI-assisted reflection, engagement with therapy, emotional awareness, therapeutic alliance, autonomy and dependency, experiences of validation, client safety, therapist observations, supervision experiences, and whether clients subsequently choose to continue into longer-term psychotherapy.
Qualitative research may be particularly valuable initially because unexpected experiences—positive and negative—need to be understood before they can meaningfully be reduced to outcome measures.
Research Should Follow the Questions
AART-345 does not begin with the assumption that AI will improve psychotherapy.
Nor does it begin with the assumption that AI will damage it.
It begins with curiosity accompanied by clinical caution.
As AI becomes increasingly present in people’s emotional lives, psychology has an opportunity—and arguably a responsibility—to study not only what these technologies can do, but what happens to people when they use them.
The development of AART-345 therefore remains open to research, professional scrutiny, collaboration and revision as evidence develops.
The purpose of research is not to prove that AI belongs in therapy.
It is to discover where it may help, where it may not, and how we can tell the difference.I think that final statement could become a particularly strong principle for the Research side of AART-345. It signals that you are approaching the subject as a psychology researcher rather than marketing AI as an established therapeutic intervention.
Acceptance, Resistance and Professional Attitudes
The integration of artificial intelligence into psychological practice is associated with markedly different attitudes among clinicians, clients and the wider public. Emerging evidence suggests that acceptance cannot be understood simply as enthusiasm for, or resistance to, technological innovation. In a large comparative study, Békés and Aafjes-van Doorn (2026) found that community participants demonstrated greater acceptance of AI-based mental-health interventions, while clinicians were more sceptical and patients occupied an intermediate position. Importantly, general attitudes towards AI were more strongly associated with acceptance than professional role or demographic characteristics.
Research specifically involving psychotherapists similarly identifies substantial variation in attitudes. Wagner and Schwind (2025) found that perceptions of usefulness and technological affinity influenced therapists’ acceptance of AI, while negative attitudes were discussed in relation to concerns about professional replacement and limited understanding of AI technologies. Forty percent of their sample described themselves as not technically inclined.
These findings suggest that resistance itself warrants investigation rather than being regarded simply as an obstacle to implementation. Professional scepticism may reflect legitimate concerns regarding therapeutic relationships, confidentiality, clinical responsibility and the preservation of human judgement. Conversely, strong enthusiasm for AI also requires critical examination. Research should therefore explore the psychological, professional and experiential factors underlying both acceptance and rejection.
77% of surveyed psychologists reported discussing patients’ AI use, 35% reported patients using AI as an additional mental-health professional, and 33% reported patients using AI to assist with therapy or treatment (APA, 2026)
Proposed AART-345 Research Questions
How do attitudes towards AI-assisted psychological reflection differ between clients, therapists, supervisors and other relevant groups?
Which factors predict acceptance, ambivalence or resistance to the use of AI within psychotherapy?
To what extent are attitudes associated with previous AI experience, technological literacy, perceived usefulness, trust and perceived risk?
What concerns underlie professional resistance—for example, client safety, confidentiality, therapeutic alliance, professional identity or fear of replacement?
Do attitudes towards AI change following direct experience of structured, therapist-guided AI-assisted reflection?
Does supervised exposure increase acceptance, increase scepticism, or produce a more differentiated understanding of where AI may and may not be clinically useful?
How do clients’ and therapists’ attitudes towards AI influence the therapeutic process and their experience of AART-345?
AART-345 does not assume that greater acceptance of AI is necessarily the desired outcome. Increased acceptance, increased caution, or a more nuanced understanding of its appropriate boundaries may each represent meaningful findings.
References
American Psychological Association. (2026). Patients are bringing AI to therapy: Highlights from the 2026 Chatbots and Mental Health Survey. American Psychological Association report
American Psychological Association. (2026, June 16). Your patients are using AI. Here’s how to talk with them about it. American Psychological Association guidance
Békés, V., & Aafjes-van Doorn, K. (2026). Who wants to have an AI therapist? Acceptance of using artificial intelligence for mental health interventions among clinicians, patients and the general community. Clinical Psychology & Psychotherapy, 33(1), e70220. https://doi.org/10.1002/cpp.70220
Malouin-Lachance, A., Capolupo, J., Laplante, C., & Hudon, A. (2025). Does the digital therapeutic alliance exist? Integrative review. JMIR Mental Health, 12, e69294. https://doi.org/10.2196/69294
Wagner, J., & Schwind, A.-S. (2025). Investigating psychotherapists’ attitudes towards artificial intelligence in psychotherapy. BMC Psychology, 13, 719. https://doi.org/10.1186/s40359-025-03071-7
World Health Organization. (2021). Ethics and governance of artificial intelligence for health: WHO guidance. World Health Organization. World Health Organization guidance
Zhang, Q., Zhang, R., Xiong, Y., Sui, Y., Tong, C., & Lin, F.-H. (2025). Generative AI mental health chatbots as therapeutic tools: Systematic review and meta-analysis of their role in reducing mental health issues. Journal of Medical Internet Research, 27, e78238. https://doi.org/10.2196/78238